Scholarlink

<p>Most existing studies provide coarse spatial resolution mappings (typically 1 km or more), which fail to capture local-scale heterogeneity of permafrost distribution in the permafrost boundary region. This study employed 298 ground-truth samples to evaluate six machine learning (ML) algorithms for simulating permafrost distribution in the Genhe River Basin (GRB) of the Greater Khingan Mountains (GKM) based on our detailed investigation (e.g., 16 boreholes) in this region conducted in 2023–2024, while identifying key environmental drivers through Shapley Additive Explanations (SHAP) analysis. Results show that the random forest (RF) model achieved the best performance, with a classification accuracy of 0.83 and a Kappa coefficient of 0.66. The RF-based permafrost map at a 30 m resolution reveals a total permafrost area of approximately 8248.5 km<sup>2</sup>, accounting for 52.0% of the GRB. The most influential predictors of permafrost distribution are slope (SLO), topographic wetness index (TWI), and degree of topographic relief (DTR), contributing 13.6%, 11.1%, and 9.4%, respectively. Other important factors include normalized difference water index (NDWI, 6.8%) and land surface temperature (LST, 6.1%). Permafrost is mainly distributed in valley bottoms, toe slopes, and gently sloping areas in the upper and middle reaches of the basin. These zones are closely associated with vegetation types such as wetlands, shrubs, and larch forests. Conversely, permafrost is rarely found in croplands or on steep slopes. These findings improve the understanding of permafrost distribution patterns in the transitional zone of Northeast China, and offer critical data and methodological support for high-resolution permafrost mapping across the region.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

High-resolution (30 m) mapping of permafrost distribution in the Genhe River Basin, Greater Khingan Mountains, Northeast China

  • 期刊论文

  • 发表日期：

    2026年3月6日

查看全文
  • Xueling Jiao,
  • Lin Zhao,
  • Defu Zou,
  • Lingxiao Wang,
  • Chong Wang,
  • Yuanwei Wang,
  • Guojie Hu,
  • Erji Du,
  • Yao Xiao,
  • Guangyue Liu,
  • Shibo Liu,
  • Yuxin Zhang,
  • Zhibin Li,
  • Minxuan Xiao

摘要

Most existing studies provide coarse spatial resolution mappings (typically 1 km or more), which fail to capture local-scale heterogeneity of permafrost distribution in the permafrost boundary region. This study employed 298 ground-truth samples to evaluate six machine learning (ML) algorithms for simulating permafrost distribution in the Genhe River Basin (GRB) of the Greater Khingan Mountains (GKM) based on our detailed investigation (e.g., 16 boreholes) in this region conducted in 2023–2024, while identifying key environmental drivers through Shapley Additive Explanations (SHAP) analysis. Results show that the random forest (RF) model achieved the best performance, with a classification accuracy of 0.83 and a Kappa coefficient of 0.66. The RF-based permafrost map at a 30 m resolution reveals a total permafrost area of approximately 8248.5 km2, accounting for 52.0% of the GRB. The most influential predictors of permafrost distribution are slope (SLO), topographic wetness index (TWI), and degree of topographic relief (DTR), contributing 13.6%, 11.1%, and 9.4%, respectively. Other important factors include normalized difference water index (NDWI, 6.8%) and land surface temperature (LST, 6.1%). Permafrost is mainly distributed in valley bottoms, toe slopes, and gently sloping areas in the upper and middle reaches of the basin. These zones are closely associated with vegetation types such as wetlands, shrubs, and larch forests. Conversely, permafrost is rarely found in croplands or on steep slopes. These findings improve the understanding of permafrost distribution patterns in the transitional zone of Northeast China, and offer critical data and methodological support for high-resolution permafrost mapping across the region.